Abstract
Keywords
The Australian National Survey of Mental Health and Wellbeing (NMHS) [1] is a landmark in the delineation of mental disorders in the Australian population. Indeed, it will probably provide the basis for mental health planning for the immediate future. One of the key findings from the NMHS was that the 12-month point prevalence of ICD-10 depression was reported to be 5.8% [1], any ICD-10 affective disorder, 7.2% [2], and any DSM-IV affective disorder, 6.6% [1]. However, even allowing for the minor differences in these results and the different time periods, they are appreciably less than a South Australian random and representative sample of urban and rural respondents who reported a two-week prevalence of symptoms consistent with DSM-IV major depression of 6.8%, with a further 10.6% having other depressions [3].
It is difficult to reconcile these differing estimates. Of course, it may be that South Australians are more depressed than other Australians, it is also possible that some of the variation was due to differences in sampling methodology, participant bias and response artefacts. Even given these, however, it is likely that important differences in the estimates remain. In seeking to explain these differences, it is pertinent to reflect on the methodologies used in the detection of depression in these reports.
The Australian National Survey was conducted by the Australian Bureau of Statistics and was of 10 641 persons (response rate 78%) in urban and rural settings across Australia. The methodology is described in detail elsewhere [2]. It used the Composite International Diagnostic Interview (CIDI) [4], administered from a laptop computer, to determine psychiatric diagnoses. The CIDI has the imprimatur of the World Health Organization, and contains several probe or screening questions to determine purported clinical significance and to exclude a number of conditions. These probe questions are designed to screen out those whose symptoms were considered to be due to medication, drugs or alcohol, or to a physical illness or injury; those who considered their symptoms to be trivial or who had not consulted a doctor; those who considered that their symptoms did not interfere ‘a lot’ with their everyday life and activity, with the definition of ‘a lot’ being determined by the respondent; and also those who had not taken medication for their symptoms on more than one occasion. Respondents who were pregnant were also excluded.
The South Australian study was of 3010 people (response rate 70%) in urban and rural settings, selected on the basis of a random and representative sample using Australian Bureau of Statistics collectors' districts. Full details are provided elsewhere [3]. Trained health research interviewers administered the mood module of the PRIME-MD 1000 study [5], from which DSM-IV depression diagnoses were determined by a computer algorithm. The PRIME-MD was developed as a screening tool for people attending primary care physicians, but it has also been used as a screening instrument in schools [6], and in a large population-based study of psychiatric disorders in pregnancy [7]. This is not unexpected, as its questions paraphrase the DSM-IV criteria for depressive symptomatology. Consistent with the full version of the PRIME-MD, two probe or screening questions about depressive symptoms experienced in the last month were also administered. However, respondents who did not endorse one of the screening questions, but who endorsed the PRIME-MD mood module items, provided inconsistent responses. Therefore, diagnoses were accepted on the basis of the DSM-IV two-week criterion for symptoms of major depression. Diagnoses were assigned at the data analysis stage, and no judgement in regard to severity or clinical significance of symptoms was made, either by the interviewer or the respondent. There was also no preclusion of a depression diagnosis on the basis that the depressive symptoms might be related predominantly to a physical illness or injury, or to substances.
While the absence of preclusion criteria using the PRIME-MD mood module would lead to an over-estimate of depression, it is doubtful whether that explains the magnitude of the difference in the results of these two surveys, and the validity of the CIDI exclusion criteria warrant further consideration.
Major depression and other depressive disorders are often associated with alcohol and/or drug use, as well as with concomitant physical illness and injury, and to exclude the depression diagnosis because of the presence of these conditions does not reflect good clinical practice. Furthermore, under the CIDI, pregnancy qualifies as a ‘physical condition that can cause symptoms’, and therefore is an exclusion criterion, although it is reassuring that the probe guidelines do acknowledge that ‘pregnancy is not a physical illness’!
The exclusion of those who consider their symptoms to be trivial also risks the omission of those who tend to deny the significance of their symptomatology and who have poor mental health literacy. Indeed, it has been demonstrated that the mental health literacy of those in the community who have major depression is no more conducive to identifying depression and recommending its treatment than for those without depression [8]. Therefore, to exclude those who believe their symptoms to be trivial is not necessarily based on existing evidence.
The exclusion of those who had thought their symptoms were sufficient to seek treatment, but who only had one course of treatment, is also liable to underestimate the prevalence of depression. The presence of sideeffects which may mitigate against further use of medication, inadequate medical management and poor mental health literacy are possible reasons why those with major depression would be excluded by this criterion.
The implication of these exclusions is that depression cannot occur as a comorbidity with a wide range of other health conditions, that it cannot occur unless it has been identified or treated over time by a medical practitioner, and that those who are in denial do not have the condition. Such exclusions are not consistent with observations that many who suffer depression do not seek medical care, or seek medical care for other comorbid conditions. Indeed, the inappropriateness of such a system of preclusion has been commented on by others, and Paykel [9] stated that the DSM-IV ‘assigns separate unjustified categories of medical and substance-induced mood disorders’.
The potential importance of such a preclusion procedure is illustrated by noting that 14.4% of those 9.9% of the South Australian sample who had asthma had symptoms consistent with the diagnosis of major depression [10], and that the exclusion of these respondents alone would decrease the prevalence of major depression by about 1%.
Henderson et al. [1] noted that the CIDI diagnoses were ‘certainly not clinicians' diagnoses’, but in response to their own enquiry of ‘are these prevalence rates believable?’, they referred to similar figures in the UK report of Jenkins et al. [11]. However, one could also refer to figures similar to the South Australian study in the comprehensive review of Angst [12]. Henderson et al. [1] also acknowledged that although ‘satisfactory procedural validity for the anxiety disorders’ had been demonstrated, ‘the research community must await more information on its validity across a range of other diagnostic groups’. In this regard it is pertinent that Rosenman et al. [13] reported that the ‘agreement between CIDI-Auto and psychiatrist on principal diagnosis was poor, Kappa = 0.23’, a finding consistent with that of Komiti et al. [14] who reported that for a major depressive episode Kappa was 0.25 between the CIDIAuto and the clinician.
It is also pertinent to reflect on a recent report from Finland, which compared CIDI diagnoses of major depression with consensus diagnoses based on the semistructured Schedule for Clinical Assessment in Neuropsychiatry (SCAN) [15]. A ‘moderate’ correspondence between instruments was found (Kappa = 0.60), and the Finnish researchers reported that ‘false negativity was mainly due to not endorsing the stem questions’ of the CIDI. It is of note that false negatives on the CIDI ‘suffered from their depression as much as those correctly identified’, and this is consistent with the observation of Andrews et al. [2] that ‘even if the prevalence figures are different, the nature of the disorders identified are likely to be the same’. That this is so is illustrated well in the South Australian data. Thus there were 18 cases who did not endorse the two probe questions, but who met the mood module criteria (i.e. DSM-IV) for major depression, and there were 186 cases who endorsed both the probe questions and who met the mood module criteria for major depression. When we compared their use of GP services (15/18 vs. 127/186) there was no significant difference between the two groups (OR = 2.32, 95% CI = 0.60–10.53). Furthermore, although none of the 18 cases were seeing a psychiatrist, they were just as likely as those with major depression to report seeing other mental health professionals such as psychologist, counsellors or community mental health workers (OR = 0.84, 95% CI = 0.09–3.97).
It is appreciated that having symptoms or a diagnosis does not necessarily mean that treatment is indicated [16]. However, among the South Australian respondents with either major or other depressions, in addition to their considerable use of health services, they experienced marked deficits in health-related quality of life, as delineated by two validated instruments [3].
It is also appreciated that there is a difference between the validity and utility of psychiatric diagnoses, and Kendell and Jablensky [17] acknowledged that ‘because utility often varies with the context, statements about utility must always be related to context, including who is using the diagnosis, in what circumstances, and for what purposes’. In this regard, although Henderson et al. [1] noted that as a result of the Australian National Survey ‘the Commonwealth, States and Territories can now make better informed changes to the way Australia's mental health can be protected and improved’, we suspect that administrators responsible for planning mental health services, and clinicians who may refer occasionally to the Australian National Survey figures, would not be aware of the exclusion criteria, discussed above, which have the potential to under-report the prevalence of mental disorders in the community.
That this has greater implications than a sterile academic debate is illustrated not only by the quality of life measures referred to previously [3], but also by considering the excess financial burden of depression calculated on the basis of loss of health-related quality of life of those with depression diagnoses in South Australia [18]. The excess burden of depression in South Australia alone was approximately $2800 million, of which 58% could be attributed to major depression and 42% to other depressions [16]. These figures are hardly trivial and are recurrent year by year. At least half this figure would not be estimated were the CIDI exclusions applied to the PRIME-MD method of classification. These data provide an indication of the effect of the probe or screening questions of the CIDI at the community level. By using the CIDI exclusion criteria and the economic assumptions in the South Australian data, and extrapolating to the total Australian population, there would be at least a $10 billion per year under-estimate of the impact of depression in Australia, with significant implications in terms of the need for resources to alleviate that burden.
We do not maintain that either method of the delineation of depression is necessarily correct. Indeed, the classification of depression is open to interpretation [19], and Kendell and Jablensky [17] have reminded us that ‘there is little evidence that most contemporary psychiatric diagnoses are valid, because they are still defined by syndromes that have not been demonstrated to have natural boundaries’. Nevertheless, we consider that because of the breadth of the exclusion criteria utilized in the probe or stem questions in the CIDI delineation of mental disorders, there is a real risk that the Australian National Survey has seriously under-estimated the extent of depression and its morbidity, both emotional and financial. The implicit corollary of this is that administrators responsible for planning mental health services may not provide adequate resources for the optimum management of depression in the Australian community.